You are on page 1of 10

Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 258

Application of Conditional Mean Spectra in Liquefaction Triggering Evaluation


Albert R. Kottke, Ph.D., P.E., M.ASCE1; Michael D. Boone, P.E., P.G., M.ASCE2;
Nick J. Gregor, Ph.D.3; and Mahi Galagoda, P.E., G.E.4
1
Pacific, Gas & Electric, San Francisco, CA. E-mail: albert.kottke@gmail.com
2
Black & Veatch, Walnut Creek, CA. E-mail: BooneM@bv.com
3
Bechtel Corporation, San Francisco, CA
4
Bechtel Corporation, Houston, TX
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

ABSTRACT
Liquefaction triggering procedures use two methods to develop the earthquake-induced
cyclic shear stress: 1) the simplified procedure and 2) site-specific response analysis. In both
approaches, site response analyses are used to calculate site-specific seismic demands. A uniform
hazard response spectrum (UHRS), which aggregates the ground motion from multiple seismic
sources, can be used as the input motion into the site response analyses. Alternatively, a suite of
conditional mean spectra (CMS) input motions can be used. In this paper, a procedure for using
CMS input motions for liquefaction triggering evaluation is proposed. The proposed procedure
uses realistically shaped input ground motions that are strongly associated with a specific
earthquake scenario, and typically results in reduced seismic demands.

INTRODUCTION
Liquefaction triggering evaluations typically relate the cyclic resistance ratio (CRR) to the
cyclic stress ratio (CSR) resulting in the factor of safety against liquefaction. The CRR is
generally based on penetration resistance or shear wave velocity. The CSR represents the
earthquake demand and is the focus of this paper.
The uniform hazard response spectrum (UHRS) quantifies the spectral accelerations with the
same return period and is used as basis for design and evaluation. The problem with using the
UHRS directly is that the UHRS aggregates ground motions from multiple seismic sources,
which produces a broadband spectral shape. This spectral shape is inconsistent with recorded
ground motions, which have energy concentrated over narrow period ranges. To address these
concerns, Baker and Jayaram (2008) proposed the concept of conditional mean spectrum (CMS).
A CMS motion matches the UHRS at one conditioning period, and the spectral accelerations at
other periods are based on the correlations observed in response spectra of recorded ground
motions. A CMS motion has a realistic spectral shape and only matches the UHRS at one period.
The use of a single CMS motion may be appropriate for systems whose response is dominated by
one mode. However, for systems whose response is dependent on multiple modes, suite of CMS
are used. The envelope of the CMS suite agrees with the UHRS and the envelope of the response
computed with CMS suite is used in the seismic evaluation.
In calculation of the CSR, the magnitude scaling factor (MSF) is used to adjust the CSR for
the magnitude effects. If a UHRS is used as the input motion, then the selecting a representative
magnitude may be complicated because of the potential for different earthquake scenarios
(magnitude, M, and distance, R) to control the hazard over different period ranges. If CMS are
used, each CMS motion is associated with a specific M and R.
Hashash et al. (2015) previously demonstrated the potential benefits of using CMS as input
into liquefaction evaluation. The Hashash et al. (2015) study considered two CMS: 0.2 s M 6

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 259

event, and 1.0 s M 7.5 event. The historical evidence in the study region showed little evidence
of liquefaction. Using the UHRS input motion, pervasive and widespread liquefaction was
predicted. CMS input motions predicted limited liquefaction and agreed with the historical data.
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

Figure 1. Measured CPT data and the site shear-wave velocity model.
In this paper, post-liquefaction settlement is evaluated for a site for ground motions with
probability of exceedance of 10% in 50 years and 2% in 50 years, 475- and 2,475-year return
periods, respectively. The earthquake demands are characterized by CSR developed through the
simplified method, as well as site response analysis. This set of analysis is performed using both
UHRS and CMS input motions. The influence on the site response, CSR profiles, and post-
liquefaction settlement are presented and discussed.

SITE CONDITIONS
For this paper, a deep soil alluvial site with Holocene, predominantly silica-based minerals is
used as a case study. The various methods of calculating CSR are applied to a single piezocone
penetration test (CPTu) sounding. The measured tip resistance, friction ratio and pore pressure
measured at the cone shoulder (u2 position) from this sounding is shown as Figure 1. This figure
indicates that sands with some silt and clay interbeds are present from the ground surface to
about 18 meters depth. Below about 18 meters depth, the materials are generally silty and clayey
with sandy interbeds. Borings, laboratory tests, and other CPTu near the exploration shown in
Figure 1 confirm these interpretations. A velocity profile was developed through a collection of
multiple SCPTu soundings and suspension logging measurements performed in the vicinity and
used to develop a site response model with site period of about 2 seconds. Nonlinear curves were
developed using index properties from laboratory tests and the Darendeli and Stokoe (2001)
model for the fine grained soil layers and Menq (2004) model for the gravelly soil layer.

DEVELOPMENT OF INPUT MOTIONS


Site specific probabilistic seismic hazard analysis (PSHA) was conducted at the site. The
deaggregation of the hazard at a return period of 2,475 years for peak ground acceleration
(PGA), 0.2 s spectral acceleration, and 2.0 s spectral acceleration is shown in Figure 2. For this

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 260

site, two scenarios primarily contribute to the hazard: a proximate moderate magnitude event and
a distant large event. The distant large event is from one fault with the same magnitude for both
considered return periods: M 8.3 at 200 km. The magnitude of the proximate moderate event
varies from M 5 to 7.5 at a distance of 30 km. For the sake of simplicity in this study, the
proximate moderate event is represented by a M 6.3 at 30 km.
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

Figure 2. 2475-yr hazard deaggregation for PGA, SA at 0.2 s, and SA at 2.0 s.

Figure 3. 475 and 2,475-yr input UHRS and CMS motions.


From the hazard curves, uniform hazard response spectra (UHRS) with return periods of 475
and 2,475 years were constructed. Selecting the representative M and R scenario for the
broadband UHRS is difficult. Given the site period of about 2 seconds, the deaggregation at 2
second is used to select the M 8.3 at 200 km as the controlling UHRS scenario for both return
periods. Here site period, not structural period, is used for the deaggreagation because the
evaluation is focused on the response of the site and liquefaction. Depending on the site
conditions and seismicity, selection a of single representative scenario for the UHRS may be
extremely challenging – although this is the state of practice.
The earthquake sources contributing to the hazard are from crustal earthquakes. Although the

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 261

PSHA was performed based on a selection of applicable crustal ground motion prediction
equations (GMPE), the results presented in this paper for this example will be based on the
Abrahamson et al. (2014) ground motion prediction equation (GMPE) to define the median and
logarithmic standard deviation (  ln ) for the two considered scenarios.
Using the UHRS and the deaggregation data, a collection of conditional mean spectra (CMS)
are developed using the methodology proposed by Kishida (2017) and the inter-period
correlations developed by Baker and Jayaram (2008). The Kishida (2017) methodology permits
selecting a range of conditioning periods, as opposed to the single conditioning period as original
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

proposed by Baker and Cornell (2006). In this study, the deaggregation data was used to select 5
period ranges: 0.01–0.03, 0.03–0.15, 0.15–0.75, 0.75–3.0, and 3.0–10.0 seconds. The period
range fully defines the UHRS and when enveloped completely matches the UHRS. As the period
range is reduced, the seismic demands are decreased, but the number of CMS motions required
increases. Selection of the appropriate number of period ranges is dependent on the analysis
methods and seismicity. The 0.75–3.0 s CMS motion is highlighted in the discussions below and
referred to as 2 s for brevity. For each period range, weight factors of the M 8.3 and M 6.3 events
are computed based on the hazard contribution.
The UHRS and 10 CMS are compared in Figure 2. The envelope of the suite of CMS
matches the UHRS, but each CMS motion only matches the UHRS over a range in periods. In
the calculation of the CMS, each scenario includes both earthquake M and R, as well as number
of standard deviations from the median (  ) and is dependent on the GMPE. As the return period
lengths from 475 to 2,475 years, the ε required to equal the UHRS at the conditioning periods
increases (indicating a lower probability event) and results in a greater distortion of the spectral shape
of the CMS motion. This distortion produces spectral shapes that are consistent with the observed
ground motions, which tend to be narrowband.
The process of partitioning a UHRS into CMS may be done using various degrees of
refinement. As the conditioning period reduces the overall demands decrease, but the number of
analyses required increases. Furthermore, the need to include multiple scenarios at each of the 5
conditioning period ranges depends on the seismicity of the site in question. For this site, it would
have been possible to use four different magnitudes at each period range.

SITE RESPONSE ANALYSES


Site response analyses were performed using the UHRS and 10 CMS input motions at the
two return periods. The analyses were performed using equivalent-linear site response analysis
with input motion characterized using random vibration theory as described in Kottke and Rathje
(2008). The analyses were conducted with Python libraries: pyrvt (Kottke 2017a) and pysra
(Kottke 2017b). Using random vibration theory eliminates the need for selection and potential
modification of earthquake time series to fit the alternative input motions.
The average response for each of the 5 CMS period ranges is computed as the weighted
average of the two considered ground motions (i.e., M 6.3 and M 8.3). The 5%-damped
acceleration response spectra at the surface of the site are shown in Figure 4. While the UHRS
defines the envelope of the CMS at the input horizon (see Figure 3), at the surface of the site
UHRS is no longer enveloping the CMS motions due to the nonlinearity of the site response
process. The UHRS induces the largest strains within the site and the lowest site amplification
(Figure 4).
It is tempting to simplify the alternative CMS spectra into one representative envelope of the
five average CMS spectra. However, if the purpose of the analyses was to provide ground

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 262

motions for use further analysis, each of the CMS motions should be considered. In this study,
the seismic demands are used to evaluate liquefaction triggering and post-liquefaction settlement.
This process ensures the maximum benefit of using CMS is included.
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

Figure 4. Surface response spectra and site amplification for 2,475-yr UHRS and CMS.

Figure 5. Influence of magnitude scale factor on the calculated CSR.


LIQUEFACTION ANALYSIS
Liquefaction analyses are often focused on the factor of safety against liquefaction and the

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 263

consequences of liquefaction. In this paper, the consequence of liquefaction evaluated is the


ground surface settlement. It is acknowledged that other consequences of liquefaction, such as
strength loss, may be important for a given project or design objective. In these cases, a similar
process could be used.

Factor of Safety Against Liquefaction


In this paper, the CRR, magnitude scaling factor (MSF), and resulting factor of safety against
liquefaction is calculated in general accordance with Boulanger and Idriss (2014). The CSR is
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

calculated using site response analysis and the Simplified Method as described by Boulanger and
Idriss (2014), as described in the following section.
The magnitude dependence of CSR/CRR is included through the magnitude scale factor
(MSF). Two MSF are used in this study: Idriss (1999) and Boulanger and Idriss (2014). The
Idriss (1999) MSF is only dependent on the magnitude, whereas the Boulanger and Idriss (2014)
depends on magnitude and soil characteristics. The Idriss (1999) MSF is used in plotting
normalized CSRM=7.5 profiles, while the Boulanger and Idriss (2014) MSF is used in the FSliq
calculation.
It is acknowledged that in practice more than one liquefaction triggering method is advised.
However, the intent of this paper is to evaluate the use of CMS in calculating CSR not a review
of liquefaction triggering methods. For this paper, the factor of safety against liquefaction is used
as input to calculate post-liquefaction volumetric strains and post-liquefaction settlement.

Cyclic Stress Ratio


The cyclic stress ratio (CSR) is used to estimate the earthquake demand through the soil
column. Generally, this is calculated either through site response or through the Simplified
Method. Using site response, the CSR is computed by Boulanger and Idriss (2014):

CSR=0.65 max' (1)
v
where  v' is pre-seismic vertical effective stress profile and  max is the site response calculated the
maximum shear stress. Alternatively, the CSR may be calculated using the Simplified Method (Seed
and Idriss, 1967):
a 
CSR=0.65 max v' rd (2)
g v
where  v and  v' are the pre-seismic vertical stress and effective stress, respectively, g is the
gravitational acceleration, and rd is the shear stress reduction factor.
When site response analysis is used to calculate amax, the site response method and the
Simplified Method result in the same CSR at the ground surface. As depth increases, the two
methods may give different values. Because the Simplified Method uses a generic shear stress
reduction factor ( rd ), whereas the site response analysis explicitly computes a site-specific shear
stress reduction factor.
Recent guidance from the National Academies of Sciences and [NAS] (2016) suggests that
because liquefaction triggering procedures were developed using the Simplified Method, and
consistent application of the triggering procedures requires use of the Simplified Method.
However, liquefaction assessments at depths beyond 10 m can benefit from site-specific CSR

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 264

profiles due to the large uncertainty in the rd at depths beyond 10 m (Boulanger and Idriss
2014). Both approaches of developing CSR profiles are used and may benefit from the
application of CMS input motions.
The CSR calculated using the 2,475-year ground motions and site response analysis are
shown in Figure 4. Figure 4a shows the CSR at the given earthquake magnitudes while Figure 4b
shows the CSRM=7.5, which have corrected to a magnitude of 7.5 using the Idriss (1999) MSF.
The highest CSR correspond to the smaller magnitude events, however when these scenarios are
scaled by the MSF, the CSRM=7.5 profiles associated with small magnitude events are reduced.
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

The magnitude corrected CSRM=7.5could then be used to compute the weighted average
CSRM=7.5 profile for the CMS motions, shown in Figure 5. The envelope of these 5 CMS profiles
would then define the CSRM=7.5 for use in liquefaction triggering. This envelope is entirely
defined by the 0.75–3 s CMS, which is less than the UHRS at all depths.

Figure 6. CSR profiles for the Simplified Method and site response.
A similar process would be used for the Simplified Method. For each CMS, amax at the
ground surface would be used to compute the CSR using Equation 2, which would then corrected
to a CSRM=7.5 using the selected magnitude. The CSRM=7.5 calculated using this procedure are
shown in Figure 5. As with the site response approach, the CSRM=7.5 computed using the UHRS
and larger than that computed by the CMS.

Post-Liquefaction Settlement
In this paper, volumetric strains and post-liquefaction settlements are calculated using the
methodology of Ishihara and Yoshimine (1992) and Yoshimine et al. (2006), as described by
Idriss and Boulanger (2008), which uses factor of safety against liquefaction and normalized
penetration resistance as inputs to calculate volumetric strain. The strains are then integrating the

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 265

vertical strains over the depth resulting in a 1-dimensional (1D) settlement profile. It is
acknowledged that alternative methods of calculating post-liquefaction ground settlement are
available and may be appropriate based on site conditions, loading conditions, and project goals.
However, the intent of this paper is to discuss the use of CMS so only one settlement method is
used for clarity.
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

Figure 7. Settlement profiles computed for the 2,475-yr motions using the Simplified
Method and site response.
Figure 7 present 1D post-liquefaction settlement profiles using the Simplified Method and
site response for the 2,475-year ground motion. This comparison indicates that using the CMS
results in reduced settlements when compared to the UHRS. The settlement calculated using the
CMS motions is the largest average settlement predicted by the 5 period ranges or the largest of
any of the CMS lines shown in Figure 7.

DISCUSSION AND CONCLUSIONS


In this study, the benefits of using CMS input motions in site response, liquefaction
triggering, and post-liquefaction settlement are demonstrated. By partitioning the UHRS into
multiple CMS, narrowband input motions with realistic spectral shapes that are directly
associated with a representative scenario (M and R) are used in the analyses. This methodology
requires the consideration of multiple input motions but provides more realistic seismic
demands.
In site response, a stronger input motion does necessarily equal a stronger motion at the
surface due to the nonlinear response of the soil column. For soil column considered, the 2,475-

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 266

yr UHRS resulted in lower ground motions at the surface than the envelope of the CMS suite for
the same return period. This is because the UHRS input motion induced the largest strains, which
attenuated the short period energy.
The site response analyses are used to calculate CSR profiles directly, or through calculation
of the surface PGA and the CSR is estimated by the Simplified Method. These CSR profiles are
used to calculate factors of safety against liquefaction and associated post-liquefaction settlement
profiles. In this study, all settlement profiles using CMS result in lower calculated ground surface
settlement estimates when compared to the use of UHRS. This difference is attributed the high
demands of the UHRS, and the complication associated with selecting the representative
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

scenario.
The use of CMS requires additional computational and engineering effort, but provides a
mechanism partitioning the UHRS into a set of realistic ground motions. Because of the
nonlinearity of the site response, the results computed by CMS should be considered more
accurate than similar results calculated using the UHRS. For sites generally controlled by a
single seismic source, the use of CMS still has the potential to reduce seismic demands because
of the influence of  on long-return period UHRS is significant. Because of the way a CMS
suite is constructed, the envelope of the suite should be used to define the design response. The
envelope should be calculated as close to the final design response as possible to take full
advantage of the CMS.

ACKNOWLEDGMENTS
The authors of this paper would like to thank Youngan Chung and Michael Lewis, both of
the Bechtel Corporation, for their input to this paper.

REFERENCES
Abrahamson, N.A., W.J. Silva, & R. Kamai (2014). “Summary of the ASK14 Ground Motion
Relation for Active Crustal Regions”. In: Earthquake Spectra 30.3, pp. 1025–1055.
Baker, J.W. & A.C. Cornell (2006). “Spectral Shape, Epsilon and Record Selection”. In:
Earthquake Engineering & Structural Dynamics 35.9, pp. 1077–1095.
Baker, J.W. & N. Jayaram (2008). “Correlation of Spectral Acceleration Values from NGA
Ground Motion Models”. In: Earthquake Spectra 24.1, pp. 299–317.
Boulanger, R.W. & I.M. Idriss (2014). CPT and SPT based liquefaction triggering procedures.
Tech. rep. UCD/CGM-14. Davis, CA.
Darendeli, M.B. & Stokoe, K.H. II (2001). Development of a New Family of Normalized
Modulus Reduction and Material Damping Curves. Geotechnical Engineering rep. GD01-1.
Hashash, Y. M. A. et al. (2015). “Conditional Mean Spectra in Site-Specific Hazard Evaluation
for a Major River Crossing in the Central United States”. In: Earthquake Spectra 31.1, pp.
47–69.
Idriss, I.M. (1999). “An Update to the Seed-Idriss Simplified Procedure for Evaluating
Liquefaction Potential”. In: TRB Workshop on New Approaches to Liquefaction. FHWARD-
99–165. Federal Highway Administration.
Idriss, I.M. & R.W. Boulanger (2008). Soil Liquefaction During Earthquakes. Earthquake
Engineering Research Institute.
Ishihara, K. & M. Yoshimine (1992). “Evaluation of Settlements in Sand Deposits Following
Liquefaction During Earthquakes”. In: Soils and Foundations 32.1, pp. 173–188.
Kishida, T. (2017). “Conditional Mean Spectra Given a Vector of Spectral Accelerations at

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V


Geotechnical Earthquake Engineering and Soil Dynamics V GSP 291 267

Multiple Periods”. In: Earthquake Spectra 33.2, pp. 469–479.


Kottke, A.R. (2017a). pyrvt. url: https://github.com/arkottke/pyrvt.
Kottke, A.R. (2017b). pysra. url: https://github.com/arkottke/pysra.
Kottke, A.R. & E.M. Rathje (2008). Technical Manual for Strata. Tech. rep. 2008/10. Berkeley,
CA.
Menq, F-Y (2003). “Dynamic Properties of Sandy and Gravelly Soils”. University of Texas
Dissertation, Austin, TX.
National Academies of Sciences, Engineering and Medicine [NAS] (2016). State of the Art and
Practice in the Assessment of Earthquake-Induced Soil Liquefaction and Its
Downloaded from ascelibrary.org by Drexel University on 10/22/18. Copyright ASCE. For personal use only; all rights reserved.

Consequences.Tech. rep. Washington, DC. doi: 1017226/23474.


Seed, H.B. & I.M. Idriss (1967). “Analysis of Soil Liquefaction: Niigata Earthquake”. In:
Journal of the Soil Mechanics and Foundations Division 93.3, pp. 83–108.
Yoshimine, M. et al. (2006). “Flow Deformation of Liquefied Sand Under Constant Shear Load
and Its Application to Analysis of Flow Slide of Infinite Slope”. In: Soil Dynamics and
Earthquake Engineering 26.2, pp. 253–264.

© ASCE

Geotechnical Earthquake Engineering and Soil Dynamics V

You might also like